Ecology of Food and Nutrition Subsistence agriculture

This article was downloaded by: [Australian National University] On: 20 May 2014, At: 15:18 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Ecology of Food and Nutrition Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/gefn20

Subsistence agriculture and child growth in Papua New Guinea I. Mueller Allen

a b

a

a

, P. Vounatsou , T. Smith & B.J.

c

a

Department of Public Health & Epidemiology , Swiss Tropical Institute , Basel, Switzerland b

PNG Insitute of Medical Research , Goroka, Papua New Guinea c

Department of Human Geography, Research School of Pacific and Asian Studies , Australian National University , Canberra, Australia Published online: 31 Aug 2010.

To cite this article: I. Mueller , P. Vounatsou , T. Smith & B.J. Allen (2001) Subsistence agriculture and child growth in Papua New Guinea, Ecology of Food and Nutrition, 40:4, 367-395, DOI: 10.1080/03670244.2001.9991659 To link to this article: http://dx.doi.org/10.1080/03670244.2001.9991659

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor

& Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.

Downloaded by [Australian National University] at 15:18 20 May 2014

This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-andconditions

Ecology of Food and Nutrition, Vol. 40(4), pp. 367-395 Reprints available directly from the publisher Photocopying permitted by license only

© 2001 OPA (Overseas Publishers Association) N.V. Published by license under the Harwood Academic Publishers imprint, part of Gordon and Breach Publishing, a member of the Taylor & Francis Group.

SUBSISTENCE AGRICULTURE AND CHILD GROWTH IN PAPUA NEW GUINEA


I. MUELLER1,2*, P. VOUNATSOU1, T. SMITH1 and B.J. ALLEN3 1

Swiss Tropical Institute, Department of Public Health & Epidemiology, Basel, Switzerland; 2 PNG Insitute of Medical Research, Goroka, Papua New Guinea; 3 Department of Human Geography, Research School of Pacific and Asian Studies, Australian National University, Canberra, Australia (Received December 9, 1999; in final form November 23, 2000)

Spatial statistical analyses of child anthropometric data were undertaken to assess the influence of systems of subsistence agriculture, in terms of staple foods and cash crops cultivated, on patterns of child growth in Papua New Guinea. These agricultural data explained between a quarter and half of the geographical variation in anthropometric growth indicators. Accounting for differences in altitude, relief and rainfall patterns, though explaining additional geographical variation, did not improve the predictions. Child growth was better in agriculture systems with cassava and sweet potato as staple crops, but worse in systems where banana, sago and taro were staple crops. Both the cultivation of all major cash crops, and sales of fish and food crops improved child growth. More intensive agricultural systems were associated with larger children indicating that the nutritional status of children benefited from intensification as well as from the introduction of cash crops into traditional subsistence systems. KEY WORDS: Child growth, child nutrition, subsistence agriculture, cash cropping, Papua New Guinea, spatial analyses

* Corresponding author. 367

368

I. MUELLER ETAL.


INTRODUCTION In traditional subsistence settings the links between environment, food production, food consumption and health are generally very close. The physical environment limits the range of foods, which can be grown, as well as influencing the distribution of important diseases, while the food production system determines the range and amounts of foods available. These interactions are crucial for understanding the nutritional situation in such settings and for assessment of the implications of agricultural change for nutrition and health (Heywood, 1987). Eighty-five percent of the inhabitants of Papua New Guinea (PNG) live in rural areas and depend on subsistence agriculture. The vast majority of both food consumed and cash incomes of rural households stem from horticulture and small holder cash cropping. The agricultural systems and the patterns of nutrition are extremely diverse. The systems range from very low intensity, shifting cultivation with long fallow periods to perennial gardens with sophisticated methods of fertility maintenance (Allen et al., 1995). The most important staples are bananas (Musa cvs), cassava (Manihot esculenta), Chinese taro (Xanthosoma sagittifolium), coconut (Cocos nucifera), sago (Metroxylon sagu), sweet potato (Ipomoea batatas), taro (Colocasia esculenta), and yams (Dioscorea sp.).

The relatively recent introduction of smallholder export cash crops such as cocoa (Theobroma cacao) or coffee (Coffea arabica & canophora) has further increased agricultural diversity. The nutritional situation of PNG is characterised by a very high prevalence of stunting (Ferro-Luzzi et al., 1978; Heywood, 1983; Heywood and Jenkins, 1992; Oomen, 1958) which has often been attributed to the low protein, zinc and energy contents of typical PNG diets (Gibson etal, 1991; Malcolm, 1970), but see (Ohtsuka et al., 1985) as well as inadequate weaning practises (Earland and Wat, 1992). Though most rural Papua New Guinean children are short in stature, striking population differences in child growth patterns were identified by the 1982/83 PNG National Nutrition Survey (NNS) (Heywood et al., 1988). The NNS was designed to obtain information on the nutritional status of children in all major environmental zones across the most of country


AGRICULTURE AND CHILD GROWTH

369

of PNG. It remains the most reliable description of child growth at a national level and provides an opportunity of comparing child growth in numerous populations using a uniform methodology. Recent small scale studies did not show any major change in nutritional status (Hide et al., 1992). Smith et al. (1991) used these data to investigate the relation of linear growth (lengthfor-age) to diet and environment and concluded that differences in diets are the main contributing factors for the pronounced differences in PNG linear growth. A reanalysis of the NNS data by Mueller and Smith (in press), which took the strong spatial patterns into account, found that variation in growth between populations could be accounted for mainly by differences in diet, while socio-economic status was the most important predictor for within population differences. In particular, good growth was associated with higher consumption of (imported) high protein/energy food such as rice, tinned fish and meat, while some local staples such as sago were associated with impaired growth. This concurrent variation makes PNG a suitable site to study the influence of subsistence agriculture on child growth. Research has concentrated on those areas identified as nutritionally problematical in the NNS, looking at the interaction between agriculture and nutrition (Hide et al, 1992 for a review). For some areas nutritional problems have been attributed to problems with local subsistence agriculture, while elsewhere other factors such as disease, migration, or environmental constraints were considered more important. A number of studies focussed on agricultural change and agricultural stress (e.g. Allen et al, 1980; Tyson, 1987; Ulijaszek et al, 1987) and the effects of the introduction of cash cropping (e.g. Groos, 1995; Heywood and Hide, 1994; Shack et al, 1990). We now present a larger scale, nation-wide investigation of the relationship of child growth to subsistence agriculture. We have linked the NNS data to the newly available Mapping Agricultural Systems of Papua New Guinea (MASP) database (Allen etal, 1995) containing detailed information on subsistence agriculture systems, and used spatial statistics to investigate: (i) the contributions of differences in staple food crops grown in

370

I. MUELLER ETAL.


agricultural systems to the geographical patterns of child growth, (ii) the food staples and cash crops which characterise agricultural systems supporting well nourished children, and (iii) the effects of intensification of subsistence agricultural systems on nutritional status of PNG children.

MATERIALS AND METHODS The NNS contains anthropometric data on children under five years of age from all but two provinces which were sampled using a complex cluster sample frame based on these environmental zones. A more detailed description of the sampling frame is given in Smith et al. (1993) and Keig et al. (1992). For the two provinces missing in the NNS, i.e. Chimbu and North Solomons Province, anthropometric data were available from provincial nutritional surveys conducted just prior to the national survey (Harvey and Heywood, 1983a; Marks, 1980). After removing children with incomplete or obviously erroneous data, and selection of a random child in families with more than one child in the survey 21,325 children were retained for the analyses. Standard normal Z-scores for the three growth indicators length-for-age (LAZ), weight-for-age (WAZ) and weight-forlength (WLZ) based on the mean PNG pattern of growth were constructed using the LMS method (Cole, 1990; Cole and Green, 1992). The rationale for this and the growth patterns thus obtained were discussed in Mueller et al. (in press). The Mapping Agricultural Systems of Papua New Guinea (MASP) database (Bourke et al., 1998) contains detailed information on crops, agricultural practices and other subsistence activities such as fishing, trading or hunting, by agricultural system, for the entire country of PNG. More than 300 different agricultural systems were defined by five attributes: crops, fallow vegetation, cropping to fallow ratios, technologies and crop and field arrangements and mapped geographically. For the present analyses differences in subsistence were characterised by information about major staples, income from cash crops and fishing, as well


371

TABLE I Distribution of children from the NNS for the agricultural variables used in the analyses


Agricultural variable

Percentage of children in category (n = 21325)

Staple crops planted Banana Cassava Chinese taro Sagoa Sweet potato Taro Yams (D. alataf Yams (D. esculenta)b

notplanted 2.1 36.7* 36.5* 56.7* 1.1 3.2 39.4* 58.1*

30% 10.6 2.4 2.3 17.6 56.2* 7.3 — 6.8

Cash crops*1 activities Betel nut Cocoa Coconutd & copra Coffee (C. arabica) Coffee (C. canophora) Pyrethrum

not income 53.2* 69.0* 74.1* 67.9 89.2* 97.0*

250 kina 2.8 5.3 7.0 — —

4.3 75.6*

70.7* 21.0

23.2 3.4

1.8 —

Food crop sales Fish sales

* levels of the most frequent highland system used as baseline in analyses. sago not planted in gardens, numbers refer to estimated contribution of sago to diet. b Combined and recoded for analysis see Figure 3 for levels. c Oil palm, an increasingly important cash crop, not included in the analyses due to major change in extension of planting in the time between the sampling of NNS and MASP data. d Coconut included only as cash crop and not staple, due to very high correlation of the two factors. a

as on the intensity of agriculture and the population densities (Table I). The data in MASP on the extent of crop plantings and on cash income are based only on subjective assessments made during rapid appraisal surveys (Chambers, 1992; Longhurst, 1981). The intensity of agriculture was quantified using the i?-value (Ruthenberg, 1980) which is an index of the duration of the cropping period in relation to the entire cultivation cycle, i.e. cropping period + fallow period. Although the NNS and MASP databases can be linked at the level of the individual village, no reliable village co-ordinates are


372

I. MUELLER ETAL.

available. Therefore we carried out spatial analyses using geographical units termed resource mapping units (RMU), which group several villages. These units were defined for planning and resource management purposes by the Division of Land Use Research of the Commonwealth Scientific and Industrial Research Organisation (CSIRO) and were the base for the sampling frame of the NNS. The RMU-based map makes it possible to link the NNS and MASP databases to the environmental data of the PNG Resource Information System (PNGRIS) (Bellamy, 1986). In order to have a continuous map the area surrounding individual RMU's was assigned to the nearest RMU via tessellation (Cressie, 1991), using an algorithm of Turner (1996). This also allowed determination of the areas which shared common boundaries, i.e. are neighbours. In order to have sufficient numbers of children for each factor level the environmental factors were recoded by pooling some of the levels used in PNGRIS (Table II for variables and classifications used). Similarly, a new classification of yam growing systems was constructed by combining the information on the planting of D. alata and D. esculenta yams contained in MASP (Figure 3 for new classification). Comparisons between agricultural systems used the most frequent combination of agricultural variables as a standard (see Table I). This combination corresponds to the most frequent system in the central highlands. The data used exhibit great variation in sample sizes within areas as well as strong spatial correlations among areas. Crude maps would therefore be subject to considerable random error, particularly in areas where the sample sizes are small. This leads to maps in which attention is drawn to those areas whose means are based on the least stable estimates. Moreover, the estimation of the effects of explanatory variables will be biased if spatial correlations are not taken into account. These problems can be overcome by using hierarchical Bayesian spatial models based on condotional autoregressive priors (Clayton and Kaldor, 1987). These models allow a spatial smoothing (based on "borrowing strength" from neighbouring regions), which is most pronounced in areas of low populations, as well as proper estimation of covariate effects at the individual and areal


373

TABLE II Environmental variables used in the analyses: Coding of factors and distribution of children according to environmental zones Factor


Altitude Relief of terrain (difference between highest and lowest point in RMU) Annual rainfall Rainfall deficit: Seasonality of climate Risk of inundation

1

Levels

Percentage of children

Low: 0-600 m(l) a Mid: 600-1200 m (2) High: > 1200 m (3+) Low: < 30 m (1,2) Moderate: 30-100 m (3) High: > 100 m (4,5)

59.8

< 3000 mm (0-3) > 3000 mm (4+) None or irregular (2-6) Regular, moderate to severe deficit (1) No (5,6) Moderate (3,4) High (1,2) No or brief flooding (0,2) Seasonal flooding (3,4) Near permanent or tidal flooding (others)

6.8

33.4 43.0 20.8 36.2 77.1 22.9 96.1 3.9

20.8 59.3 19.9 73.7 17.9 8.4

Numbers in brackets denote original PNGRIS classifications (Bellamy, 1986).

level. Such models have been successfully applied to model extra-Poisson variation in disease risk and mortality data (e.g. Bernadinelli and Montomoli, 1992; Mollie, 1995) and recent models for normal data (Gelfand et ah, 1998) have been adapted by Mueller and Vounatsou (1999) to the context of nutritional ecology. Following common Bayesian practice the results for variables of interest are reported as median and 90% credible intervals of posterior densities. The empirical among area variance is used as a summary measure of spatial variation present in the data. The model choice is done using a Bayesian criterion developed by Gelfand and Gosh (1998). A more detailed description of the statistical modelling and inference is given in the appendix.

374

I. MUELLER ETAL.


RESULTS Child growth in PNG shows strong regional differences, as indicated by the considerable spatially structured variation in all three growth indicators (see Table III). The spatial patterns were strongest in weight-for-length (between area variance 0.184) and least in weight-for-age (0.147). The observed spatial patterns of child growth are displayed in Figure 1. The inclusion of data from Chimbu and Bougainville Provinces did not affect spatial patterns of growth described by Mueller et al. (in press). Children with the most impaired growth status were found in lowlands (i, vi), some islands (vii) and in highland fringe areas (e.g. ii), while the best growth was observed in the PNG Islands TABLE III Model comparison and posterior medians and 90% credible interval between area variance Model

Spatial varianceb

Goodness of model fit (D°o)

of

a

Median

HAZ (i) Spatial structure only (ii) + Staples & cash crops (iii) + Staples & cash crops + environment WHZ (i) Spatial structure only (ii) + Staples & cash crops (iii) + Staples & cash crops + environment WAZ (i) Spatial structure only (ii) + Staples & cash crops (iii) + Staples & cash crops + environment

35,584 35,526 35,529

0.164 0.128 0.108

[0.154,0.175] [0.114,0.146] [0.096,0.124]

35,073 35,044 35,053

0.184 0.102 0.096

[0.174,0.195] [0.089,0.119] [0.084,0.111]

36,095 36,001 36,021

0.147 0.110 0.102

[0.137,0.157] [0.096,0.129] [0.088,0.118]

a Model choice based on criterion by Gelfand and Gosh (1998), smaller values indicated better model fit. b measured as empirical variance of spatial effects: big values indicate large between area variation.

375



FIGURE 1 Observed spatial patterns of mean child growth in PNG: A— Length-for-Age, B—Weight-for-Length; C—Weight-for-Age. Posterior medians of spatial effects: • < -0.45; B -0.45-0.16; B -0.15-0.15; • 0.16-0.45; • >0.45 in SD from national mean; • Area not sampled in NNS. Specifically identified areas: (i) Torricelli/Prinz Alexander Range, (ii) Anga area, (iii) Western Province, (iv) PNG Islands region, (v) Central Highlands, (vi) Madang Province, (vii) Milne Bay Province, and (viii) Pomio District.


376

I. MUELLER ETAL.

regions (iv, except viii). Children from the central highlands (v) were heavier, those from Western Province (iii) taller, but slimmer than average. Differences in subsistence agriculture accounted for a considerable part of the geographical patterns of child growth (Table III). The geographical variation, as measured by the empirical variance of the spatial effects